What 76 AI-Scored Makeup Looks Tell Us About Application, Color, and Finish

Makeup scores differently from outfits and I keep having to remind people of that. An outfit gets graded on fit, color, occasion, styling, and details. A makeup look gets graded on base finish, color theory, application technique, feature enhancement, and occasion. Same overall 0–100 ceiling. Different game underneath.

Over the last 30 days, the platform has scored 76 makeup submissions. The average overall makeup score is 72 out of 100. That number tends to run slightly tighter than outfit scores, which makes sense: makeup is harder to get spectacularly wrong than an outfit, and also harder to get spectacularly right. The variance is compressed. Most looks land within ten points of each other.

Here's what the rest of the numbers say.

The makeup curve is tighter than the outfit curve

The score distribution for makeup is narrower than for outfits. Fewer 90+ scores. Fewer scores below 60. More of the curve sits in the seventies. That's a structural artifact of the scoring rubric: makeup has fewer obvious "fails" because most people who bother with makeup at all have a working baseline. The penalty band that wrecks outfit scores — bad fit — has no direct equivalent on a face.

The 90+ tier is small because the rubric defines "flawless" strictly. To land there, every one of five sub-scores has to clear 18/20, and most submissions have at least one dimension that drops to 15 or 16. Usually it's the same one. We'll get to that.

Which makeup dimension is pulling the average down

Five sub-dimensions, each out of 20. Ranked by current platform average, lowest at the top:

Base & Skin Finish is the lowest right now at 16/20. Occasion Appropriateness is leading at 17/20. Five dimensions sitting that close together means the difference between an 82 and a 76 is rarely a single failure — it's usually one or two sub-scores that quietly underperform across multiple submissions.

Base and skin finish is the lowest-scoring dimension, and that tracks with what we see in submissions. Foundation that doesn't match the neck. Concealer that oxidizes warmer than it looked in the pan. Powder set so heavily it kills the natural light hitting the cheekbone. Base is also the most photo-sensitive dimension — uneven lighting in the submitted photo can drop the score a point or two even when the actual application is solid. It's worth shooting twice: once in window light, once in overhead.

The skin-finish ceiling

If I had to name the single dimension that pulls every other makeup score up or down with it, it's base. Base finish is the foundation literally and figuratively: if the skin looks like skin, the rest of the look has somewhere to sit. If the base reads as cakey or mismatched, even excellent eyeshadow scores lower because the eye is unconsciously comparing it against the off-tone face surrounding it.

Looking at the high-scoring tier specifically, the looks that crossed 85 almost all have base scores at 18 or above. That correlation is tighter than for any other dimension. A look can land in the 80s with mediocre color theory or with no real feature enhancement, but it almost never crosses 85 with weak base.

Translation: if you're going to spend ten extra minutes on one part of the routine, spend them on the base. Match-test the foundation in window light, not bathroom light. Prep the skin properly first. Set strategically, not universally. None of this is news to anyone who's been doing makeup seriously. It just keeps showing up in the data as the difference between "good" and "very good."

Why outfits and makeup don't share a rubric

Every time we get asked why the outfit score and the makeup score weight differently in a complete-look analysis, the answer comes down to what each does in a photograph. Outfits define silhouette, color volume, and context cues. Makeup defines face proportion and visual focus. The rubric weights outfit at 50%, makeup at 30%, and accessories at 20% of the overall complete-look score, and that ratio is intentional — it reflects how much each contributes to how the whole look reads at a glance.

Practically: if you're looking at a low complete-look score and trying to figure out where to focus, the outfit is the biggest lever by raw weight. But makeup scores are usually the one with the most room to improve quickly, because the fixes are technique-based rather than wardrobe-based. You can change your foundation match tomorrow. Changing how the cut of a blazer reads on your shoulders takes a tailor.

What the 85+ tier consistently has

Going through the top makeup submissions, three things keep showing up:

• The base is matched to the chest, not the wrist. Foundation shopping rule that does more for scores than any other single fix. The face and the chest are visible in most photos. They need to be the same color.
• Eye and lip aren't both competing. One feature is doing the talking, the other is supporting. A bold lip with a neutral eye scores higher than a strong eye plus a strong lip, almost always.
• The finish matches the lighting. Dewy in soft natural light scores beautifully. Dewy under overhead bathroom fluorescents reads as oily. Matte is the reverse. The submitted photo's lighting is part of the score, even if it shouldn't be.

None of this is technique you can't learn. It's just technique you have to actually apply, in the right order, with the right product. The platform is reasonably good at telling you which of the three you're missing — that's most of what the dimension breakdown is for.

How the makeup score actually works

Same engine as the outfit score: Google's Gemini, prompted with a fixed five-dimension rubric, each dimension scored out of 20 using a five-tier scale from 0–6 (severe issues) up to 18–20 (flawless). The five sub-scores sum to the 0–100 overall score. No bonus for repeat users, no penalty for first-time uploads. Each look is judged against the rubric, not against your previous submissions.

Aggregate data on this page comes from the live analyses table. No images. No faces. No usernames or device IDs. Just scores, the occasion the user set before submitting, and timestamps. The window rolls every twenty-four hours.

Looking for outfit data instead? The outfit trends report is here.